Monday 22 June 2020

Intel’s Tiger Lake Mobile CPUs May Launch Next Month

We know that Intel is planning to launch a refresh of its Ice Lake mobile platform in the not-too-distant future. Data from resellers suggest the CPUs could debut as soon as next month.

Dutch website CentralPoint has two Tiger Lake-equipped Acer laptops expected to be up for sale on July 27. The two Acer Swift 5 systems reference a Core i5-1135G7 and a Core i7-1165G7, both of which offer 16GB of RAM and a 1TB SSD.

CentralPoint

CentralPoint listing.

If rumors are accurate, Intel’s 11th Gen family of Tiger Lake parts, based on the Willow Cove CPU architecture, will still be limited to 4C/8T configurations. Not much is known about the CPU’s underlying architecture, but it’s expected to deliver additional IPC and frequency uplifts relative to Ice Lake. Intel’s Ice Lake is significantly more efficient than the five-year-old, Skylake-derived, Comet Lake architecture currently deployed on desktops, but Intel had to give up significant clock speed to bring the chip to market.

Tiger Lake is expected to win some of that frequency back, though details are unclear. The CPU will be built on Intel’s 10nm++ process, which is expected to further improve the implementation of the troubled process node.

On the graphics side, Tiger Lake will be the first Intel product to debut the Xe graphics architecture. While Intel has developed its own integrated graphics for many years, this time around we’ll be seeing the first implementation of an architecture Intel intends to scale up to discrete levels. Expect a lot of low-level performance analysis with an eye towards how effectively the GPU might or might not scale. We expect to see Tiger Lake debut up to 96 execution units (EUs), which is a 1.5x increase over Ice Lake.

Intel might take the clock speed of its iGPU lower to compensate for width — this is a common practice — but assuming its memory subsystem can keep up with bandwidth pressure, we’d expect very strong scaling from adding additional cores. 96 EUs is expected to work out to 768 shader cores, though we recommend against attempting to compare AMD versus Intel GPUs on core counts. Leaked results have suggested the Xe-based GPU might be slightly faster than its AMD counterpart in Time Spy, though this is just a single unconfirmed data point.

Intel taking the lead or substantially competing with AMD in the integrated graphics arena would be a major event. It could impact gamer perceptions of the discrete cards when those appear. Personally, I’m curious to see how Intel’s solution stacks up to AMD’s as far as memory bandwidth efficiency.

Integrated GPUs live and die by memory efficiency. The one constant of integrated graphics from the introduction of the capability straight through to the present day has been memory bandwidth pressure. There have been various strategies for solving the problem, from on-motherboard dedicated RAM chips to Intel’s 128MB EDRAM pool for various flavors of Iris graphics. Obviously factors like driver support and efficient GPU cores are also important — but assuming Intel is able to deliver on these fronts, memory bandwidth will be the bottleneck and we’ll see which company can handle the problem better.

Now Read:



No comments:

Post a Comment